This manuscript reports quadruple sulfur isotope measurements using the MAT 253 Ultra high resolution IRMS and suggests that a higher D36S precision (2-5 folds) can be achieved compared to measurements using traditional IRMS such as 253 plus. The high resolution of Ultra is definitely useful to eliminate the possible isobaric interference of mass 131 (36SF5+) by C3F5+ or tungsten, and this work is an important contribution and should be eventually published in RCM.

However, my major concern is that the conclusion of this manuscript (improve D36S measurement precision by 2 to 5 folds) may be overstated (or mistakenly presented). Based on my own experience in two different labs (Thiemens Lab at University of California San Diego and my lab at Guangzhou Institute of Geochemistry, Chinese Academy of Sciences), it is not difficult to get the D36S standard deviation of around 0.05 per mil reported here for multiple zero-enrichment measurements if the voltage of mass 127 is >2000 mV. We can also achieve a similar D36S standard deviation of around 0.05 per mil for overall measurements starting from the silver sulfide standard IAEA-S1 if the sample size is sufficiently large (though the result is based on the condition of the fluorination line such as the purity of BrF5). For example, our recent batch of IAEA-S1 fluorination and measurements (3-11 umol) show a D36S standard deviation of 0.043 per mil (n=5). Therefore, I am not fully convinced that Ultra would be a perfect solution for high PRECISION D36S measurements if more advantages are not provided or discussed. Did the authors made such measurements using a small sample size? Please clarify the sample size introduced into the mass spec or the voltage of mass 127 (32SF5+). In addition, the integration time is 33s for each cycle and 50-60 cycles are needed for one measurement. The required measurement time for one sample looks quite long compared to measurements using the traditional IRMS (e.g., 253 plus) method, in which we usually do <20s integration time for each cycle and only 10-20 cycles. If a significantly higher precision cannot be achieved, is such a long measurement time really needed?

From my point of view, the major advantage of the method developed by the authors is that we are confident that any possible isobaric interference of mass 131 (36SF5+) can be fully eliminated. This is crucial for measuring natural samples that we do not know the true value. For traditional measurements using 253 plus, we don’t have strong evidence that measured D36S is true because trace amounts of carbon may lead to a wrong value and we may not identify such contamination by checking other masses such as 12 (C+), 69 (CF3+) or 119 (C2F5+). Using the Ultra as shown in this work, we can easily identify and eliminate isobaric interference, and this leads to a high accuracy. Therefore, I suggest the authors highlight the advantage of Ultra as ACCURACY improvement rather than PRECISION. Overall, I will be happy to see this work published in RCM, if my major concern outlined above can be addressed.

I listed my other comments as follows.

L41: Reference #5 (Lin et al., 2018) argued that D36S anomalies in tropospheric sulfate aerosols originate from chemical reactions in combustion rather than stratospheric input.

L43-47: I am not sure if this is the best example. If there is a magnetic isotope effect (MIE), we will see D33S anomalies and zero D36S. In the authors' writing, it is not clear how high-precision D36S measurements are important for identifying MIE. Please rewrite and further clarify. It may be written as “the near-zero D36S values led the authors…“.

L67: Ar is a major carrier gas used by MC-ICP-MS and 36Ar hinders 36S measurements. 36S may be measured by MC-ICP-MS if a collision cell is used. The references (#16, #17) cited by the authors only measured 32S, 33S and 34S.

L69-71: There are many reported D36S data (mainly for Archean samples and some for meteorites) measured by SIMS.

L139: In the introduction it was mentioned that the resolution is up to c.a. 15000 (L104). Please clarify.

L142-143: Please explain which parts are automated and which parts are not.

Mang Lin
linm@gig.ac.cn

I thank the authors for constructively revising their manuscript.

Regarding the precision issue discussed in the response and revised manuscript, I want to make one more minor comment. It’s reported in the abstract and conclusion that the D36S SD is 0.069 per mil (n=8) for IAEA S1, but please note that for IAEA S2 and S3, your D36S SD is 0.13 and 0.10 per mil, respectively (Table 2 in the revised manuscript). I am not sure if it is a good idea to only report the smallest SD (but to ignore larger ones) in the abstract and conclusion.

There are some papers reporting relatively small D36S standard deviations (similar to this work) for multiple Ag2S measurements (including fluorination, purification, and measurements by MAT 253 or 253 plus). For example, in the paper by Lin et al. (2018; https://doi.org/10.1073/pnas.180342011), if you check the raw data in the supporting information (Table S2 in that paper), both UCSD and USTC labs show a SD of c.a. 0.07 per mil (n=5 in each lab) during the study. Though in the main text of that paper, we preferred reporting errors of 0.2 per mil based on all data from years of Ag2S standard measurements (including many “low-quality” data after bad fluorination). We all know that there are uncertainties for real samples, so it would be safe to report a large SD.

The use of Ultra can clearly eliminate possible contamination as noted in the manuscript. The accuracy improvement itself is a big achievement. I like the other reviewer’s comment that “D36S scale has not been confirmed in a VCDT scale”. Ultra would be useful. Regarding precision, if a long-term monitoring of Ag2S standard fluorinations and measurements (rather than only 8 trials in this manuscript) still shows a low SD, I will be more convinced that a high precision may be achieved by using Ultra. Otherwise, the long-standing D36S precision problem may be attributed to other factors such as small fractionation during fluorination.

I will let the authors decide if they want to mention that the D36S SD observed in IAEA S2 and S3 fluorinations are larger than IAEA S1 in the manuscript, especially in the abstract and conclusion. Because not all readers may read the manuscript and check all data carefully, ignoring those larger SD may be misleading. As an analytical chemist, I will prefer to be cautious.

Overall, it is a great work. I recommend its publication in RCM.

Mang Lin
linm@gig.ac.cn